Skip to main content

Taco Cohen: Group Equivariant Convolutional Networks

Movies Preview

texts
Taco Cohen: Group Equivariant Convolutional Networks


Published June 27, 2016
SHOW MORE


Talk by Taco Cohen from the University of Amsterdam.  Given to the Redwood Center for Theoretical Neuroscience at UC Berkeley.

Abstract
The Group equivariant Convolutional Neural Network (G-CNN) is a new kind of neural network that obtains better sample complexity by exploiting symmetries. G-CNNs use G-convolutions, a new type of layer that enjoys a substantially higher degree of weight sharing than regular convolution layers. G-convolutions increase the expressive capacity of the network without increasing the number of parameters, are easy to use, and can be implemented with negligible computational overhead for discrete groups generated by translations, reflections and rotations. G-CNNs achieve state of the art results on CIFAR10 and rotated MNIST. In this talk we will briefly discuss the basics of symmetry groups and their relation to computer vision and machine learning, and then present the theory and practice of G-CNNs in a mathematical and visual manner. No prior knowledge of group theory is required.


Language English
Collection opensource_media

comment
Reviews

There are no reviews yet. Be the first one to write a review.
SIMILAR ITEMS (based on metadata)
Arxiv.org
by Joan Bruna; Arthur Szlam; Yann LeCun
texts
eye 26
favorite 0
comment 0
Source: http://arxiv.org/abs/1301.3537v1
Arxiv.org
by S. Estrada; J. R. Garcia Rozas; J. Peralta; E. Sanchez Garcia
texts
eye 22
favorite 0
comment 0
Source: http://arxiv.org/abs/math/0702543v1
Arxiv.org
by Anton Alekseev; Eckhard Meinrenken; Chris Woodward
texts
eye 16
favorite 0
comment 0
Source: http://arxiv.org/abs/math/9905130v1
Community Media
image
eye 4
favorite 0
comment 0
Community Media
image
eye 4
favorite 0
comment 0
Community Media
image
eye 12
favorite 0
comment 0
Community Media
texts
eye 144
favorite 0
comment 0
Community Media
texts
eye 20
favorite 0
comment 0
Arxiv.org
by Giuliano G. La Guardia
texts
eye 18
favorite 0
comment 0
Source: http://arxiv.org/abs/1212.4653v2